Research and Validation of Potato Late Blight Detection Method Based on Deep Learning

Late blight, caused by phytophthora infestans, is a devastating disease in potato production. In severe cases, this can lead to crop failure. To rapidly detect late study, deep learning model was developed discriminate the degree of leaf diseases with high recognition accuracy and fast inference speed. It constructed total seven categories datasets single complex backgrounds, which were augmented using data enhancement method increase number images 7039. performance pre-trained for fine-grained classification evaluated comprehensively terms accuracy, speed, parameters. The ShuffleNetV2 2× better generalization ability faster speed selected improved. Three improvement strategies proposed: introducing an attention module, reducing depth network, 1 × convolutions. Their effects on underlying explored through experiments, best form determined. loss function improved converged 0.36. This compared base model, reduced 34.5%. meantime, parameters, FLOPs, size approximately 23%, increased 0.85%, CPU 25%. Deploying embedded device, overall precision 94%, average time taken image 3.27 s. provided critical technical support automatic identification blight.